Diptanshu singh

Flow control is a fundamental concept in the design of high-performance networks. In part three of this three-part deep dive series, @networkporter explores how flow control is implemented in switch ASICs. blog.apnic.net/2025/11/07/flo…

The bottleneck for deep skill isn't usually intelligence, but boredom tolerance. Learning has an activation energy: below a certain skill threshold, practice is tedious, but above it, it becomes a self-sustaining flow state. The entire battle is persisting until that transition.

In this guest post, @networkporter shares notes on flow control, beginning with a deep dive into the principles of flow control, including key models, typical schemes, and system-level considerations. blog.apnic.net/2025/10/15/pri…

AQM notes: EWMA choices, D-ECN, and the trap of “average” depth. The math matters, but understanding your actual traffic patterns matters more. dipsingh.github.io/Active-Queue-M…

GenAI isn't just a technology; it's an informational pollutant—a pervasive cognitive smog that touches and corrupts every aspect of the Internet. It's not just a productivity tool; it's a kind of digital acid rain, silently eroding the value of all information. Every image is no longer a glimpse of reality, but a potential vector for synthetic deception. Every article is no longer a unique voice, but a soulless permutation of data, a hollow echo in the digital chamber. This isn't just content creation; it's the flattening of the entire vibrant ecosystem of human expression, transforming a rich tapestry of ideas into a uniform, gray slurry of derivative, algorithmically optimized outputs. This isn't just innovation; it's the systematic contamination of our data streams, a semantic sludge that clogs the channels of genuine communication and cheapens the value of human thought—leaving us to sift through a digital landfill for a single original idea.

Being brilliant along one axis (e.g. programming) does not give you a license to have your bad behaviour towards others tolerated. I've been fortunate to receive critical but kind feedback from utterly brilliant engineers, delivered in a direct manner but with grace. This made me respect these engineers even more, admiring them not only for their technical achievements, but also for the magnanimous way they held themselves in front of others, prioritizing mentorship over throwing their weight around. Be kind, strive for understanding and empathy.

Some notes on flow control. I revisited core ideas like XON/XOFF vs credit-based flow control (CBFC), the role of RTT in buffer design, ASIC-level trade-offs, and a quick look at M/D/1 queues and tail bounds. dipsingh.github.io/FlowControl-Hi…

Notes on Packet Classification cover theoretical and practical implementation approaches. I find this area mildly boring, but it's a hard problem space with an N-dimensional headache. dipsingh.github.io/HighSpeed-Pack…

My notes on hashing in the context of packet processing: dipsingh.github.io/Hashing-HighSp…

You may know Hamming Code, but here is Hamming Wisdom youtube.com/clip/Ugkxe4dV6…

Network traffic modelling approaches | APNIC Blog blog.apnic.net/2025/01/29/net…

One of my goals this year is to invest in translations for 3blue1brown. I’d like to hire translators directly, preferably ones with experience teaching, and who are willing to help experiment with some software tools which I hope can make the process less tedious. Full job posting: docs.google.com/document/d/1co… Application: forms.gle/9XwpNXSEHt8LCA… For early experiments, we’re budgeting for five languages (Arabic, French, German, Spanish, Hindi). If you’re interested in helping out with other languages, feel free to apply, and if we have the capacity to hire more or coordinate volunteer efforts, we’ll let you know. -- Relatedly, last year my friend @beneater put together a site to help coordinate text translations for videos. YouTube used to have a feature like this, but it was discontinued. I’m deeply grateful to everyone who has helped submit translations there, it’s not only directly useful for subtitles, but it’s a helpful first step for any dubs. criblate.com -- A number of companies, including YouTube, have been pushing AI dubs. Many of them generate a voice that sounds like the original narrator, which is a bit mind-blowing at first, but my overall impression is that once you get past that first wow factor, AI dubs are simply not that engaging. I would much rather the voice of the channel in Spanish (or any other language) be a real voice that conveys the feeling of a friendly math teacher than have it be something that vaguely sounds like me.

I spent some time learning about transistors. “No exponential is forever, but we can delay ‘forever’.” — Gordon Moore. dipsingh.github.io/Transistor-Adv…

@door7302 @networkservice @thatJuniperGuy That would be awesome.

@networkporter @networkservice We could do the same for PTX, this something we can plan with @thatJuniperGuy

love me a life of a packet blog post. kickass @door7302 ! community.juniper.net/blogs/david-ro…

@door7302 @networkservice Somehow, I missed this. Now I need to read it. @door7302 Is there one already, or is a similar one planned for PTX?

@networkservice Thank you ☺️

Me

Who is interested in Kalman filters?

@netldn Possible to get hold of slides/recording ?

It's #NetLdn 57 this Thursday, and we've got two great talks lined up, followed by some interesting discussion I'm sure:

@jwbensley Absolutely , you are interviewing them (passively at least) as much as they are interviewing you.

@networkservice @stubarea51 You have to tell me more about Sonic. Is it that bad?

@stubarea51 if sonic was a lego set most of the pieces wouldn’t pass the QA tolerance check

If #openstack was a Lego set, it would come with 1000 of the required 2000 pieces and be missing every other page in the instruction manual which you have to download via PDF from 17 different locations. But otherwise it's a complete set, ready to build